Energy Blog: AI and Energy are Intertwined
Artificial intelligence is virtually inescapable. Companies are embedding their AI-enabled tools on web searches, word processing programs, and the home screens of phones, and AI text and images are crowding out human-produced content online and elsewhere. In addition to consumer-facing products, machine learning and large language models (which are the technological underpinnings of what is broadly referred to as “AI”) are helping scientists and engineers conduct research and mine useful signals from noisy data.
Those models and algorithms require computer time to run—and demand greater resources than conventional computing. According to one estimate, a search conducted using AI tools is 10 times more energy intensive than a simple Google search (and isn’t always more reliable).
The projected growth in AI usage has led to a realization that energy demand may well explode over the next few years as more consumers use AI to accomplish everyday tasks.
The International Energy Agency, an intergovernmental organization based in Paris, released a report on the connection between energy and AI earlier this month. Called simply, Energy and AI, the 300-page document was intended to provide policymakers with up-to-date forecasts on electricity demand associated with AI, as well as knock-on impacts with regard to security, energy prices, and emissions.
The homepage for the report includes an AI chatbot to help navigate the report, when asked for the key takeaways, its answers were so general to be useless. Instead, here are five key points as determined by an actual human:
Energy demand could expand dramatically. Data centers already account for about 1.5 percent of the world’s electricity consumption, and individual AI-focused data centers draw as much electricity as 100,000 households. In five years’ time, the IAE projects that electricity demand could more than double, reaching 945 TWh per year in 2030. That’s roughly on par with the total electric demand of Japan. That growth will be disproportionately concentrated in the United States. “In the United States, data centers account for nearly half of electricity demand growth between now and 2030,” the report states. “By the end of the decade, the country is set to consume more electricity for data centers than for the production of aluminum, steel, cement, chemicals and all other energy-intensive goods combined.”
Renewables can meet much of that demand. The projected demand from data centers has been used as a rationale for recent moves by the Trump Administration to support coal power in the United States. According to an April 8 executive order, “Our Nation’s beautiful clean coal resources will be critical to meeting the rise in electricity demand due to the resurgence of domestic manufacturing and the construction of artificial intelligence data processing centers.” The IEA report sees things differently, projecting that renewables can meet nearly half the growth in data center electric demand between now and 2030, and nuclear power can satisfy another 10 percent. Even in the United States, which has less renewable power on the grid (and in the development pipeline) than many other countries have, most of the new AI demand is expected to be met by natural gas rather than coal.
Energy availability could limit AI growth. At present, AI-oriented servers are predominantly found in the U.S., especially in clusters around Northern Virginia, North Texas, Chicago, Omaha, and the Pacific Northwest. That means that AI-related electricity demand is extremely lumpy. “There are 6 states in the United States where data centers already consume over 10 percent of the electricity supply, with Virginia leading at 25 percent,” the report states. The grid is already strained in some areas due to these data centers, and future growth in those areas could be limited or other priorities, such as affordability or manufacturing growth, may have to be sacrificed.
AI could boost energy production. While electric demand from data centers is a challenge, the technology could become a boon to the industry if used correctly. The IEA report identified several avenues, such as improved petroleum drilling or increased efficiency in grid operations, where AI could give back much of the energy its data centers draw. AI-enabled building systems, for instance, could save as much as 300 TWh per year, which is more than a third of the technologies projected electric demand by 2030.
While it can help, AI won’t solve climate change. Some techno-optimists posit that unleashing AI will unlock so many efficiencies and new breakthroughs that not only will it be a net gain in terms of climate action, it could well provide a definitive solution. While that could possibly come to pass, that’s not what the IEA is forecasting. “We estimate that emissions reductions from the broad application of existing AI-led solutions to be equivalent to around 5 percent of energy-related emissions in 2035,” the report states. That makes AI part of a solution, but it doesn’t obviate the need for other actions. And some of the applications advanced and widespread AI makes possible, such as autonomous vehicles, could increase emissions if it leads passengers away from public transportation.
The entire report can be downloaded at the IEA website.
Jeffrey Winters is editor in chief of Mechanical Engineering magazine.
Those models and algorithms require computer time to run—and demand greater resources than conventional computing. According to one estimate, a search conducted using AI tools is 10 times more energy intensive than a simple Google search (and isn’t always more reliable).
The projected growth in AI usage has led to a realization that energy demand may well explode over the next few years as more consumers use AI to accomplish everyday tasks.
The International Energy Agency, an intergovernmental organization based in Paris, released a report on the connection between energy and AI earlier this month. Called simply, Energy and AI, the 300-page document was intended to provide policymakers with up-to-date forecasts on electricity demand associated with AI, as well as knock-on impacts with regard to security, energy prices, and emissions.
The homepage for the report includes an AI chatbot to help navigate the report, when asked for the key takeaways, its answers were so general to be useless. Instead, here are five key points as determined by an actual human:
Energy demand could expand dramatically. Data centers already account for about 1.5 percent of the world’s electricity consumption, and individual AI-focused data centers draw as much electricity as 100,000 households. In five years’ time, the IAE projects that electricity demand could more than double, reaching 945 TWh per year in 2030. That’s roughly on par with the total electric demand of Japan. That growth will be disproportionately concentrated in the United States. “In the United States, data centers account for nearly half of electricity demand growth between now and 2030,” the report states. “By the end of the decade, the country is set to consume more electricity for data centers than for the production of aluminum, steel, cement, chemicals and all other energy-intensive goods combined.”
Robot, M.D.
Today's surgical robots are more capable than ever before, and engineers are pushing the boundaries even farther.
Energy availability could limit AI growth. At present, AI-oriented servers are predominantly found in the U.S., especially in clusters around Northern Virginia, North Texas, Chicago, Omaha, and the Pacific Northwest. That means that AI-related electricity demand is extremely lumpy. “There are 6 states in the United States where data centers already consume over 10 percent of the electricity supply, with Virginia leading at 25 percent,” the report states. The grid is already strained in some areas due to these data centers, and future growth in those areas could be limited or other priorities, such as affordability or manufacturing growth, may have to be sacrificed.
AI could boost energy production. While electric demand from data centers is a challenge, the technology could become a boon to the industry if used correctly. The IEA report identified several avenues, such as improved petroleum drilling or increased efficiency in grid operations, where AI could give back much of the energy its data centers draw. AI-enabled building systems, for instance, could save as much as 300 TWh per year, which is more than a third of the technologies projected electric demand by 2030.
While it can help, AI won’t solve climate change. Some techno-optimists posit that unleashing AI will unlock so many efficiencies and new breakthroughs that not only will it be a net gain in terms of climate action, it could well provide a definitive solution. While that could possibly come to pass, that’s not what the IEA is forecasting. “We estimate that emissions reductions from the broad application of existing AI-led solutions to be equivalent to around 5 percent of energy-related emissions in 2035,” the report states. That makes AI part of a solution, but it doesn’t obviate the need for other actions. And some of the applications advanced and widespread AI makes possible, such as autonomous vehicles, could increase emissions if it leads passengers away from public transportation.
The entire report can be downloaded at the IEA website.
Jeffrey Winters is editor in chief of Mechanical Engineering magazine.
As renewable energy gains momentum, oil and natural gas continue to remain primary global energy sources. This collection delves into latest digital trends enabling breakthroughs in energy transformation.